Packet.ioData infrastructure for power operators

Build analytics, automation, and controls on site. We handle the data infrastructure.

For teams building applications on power assets. Packet turns raw site data into one validated asset model, then deploys and runs your applications on every site.

See the product
Packet appliance and the Map & Contextualize view, mapping raw tags to the canonical asset model
Built for
  • Battery storage
  • Solar
  • Wind
  • Substations

01Problem

Building on site data is hard. Scaling it across sites is harder.

Before any application can run, someone has to unify the data, build the runtime, and manage it on every site.

  1. 01

    Raw data has no context

    Raw tags like INV3_ACPWR_L1 carry no unit, asset identity, or hierarchy. Building a central asset model means establishing what each tag means, tag by tag, manually.

  2. 02

    Applications are rebuilt per site

    Applications must be re-architected for each hardware target, with buffering, feature engineering, and offline recovery solved from scratch.

  3. 03

    Sites are managed one at a time

    Patching, certificate renewals, and updates are handled one node at a time, so cost scales linearly with fleet size.

  4. 04

    Fleets are growing

    Mixed equipment vendors (OEMs) across sites reduce visibility and raise management overhead.

  5. 05

    Labour is constrained

    The specialized site engineers needed to maintain fleets are few and already fully utilized.

One site is possible. A fleet is not.

02Applications

What runs on Packet.

Once the data is modelled and the runtime is in place, any application that should run on site can.

Interprets data

Analytics

AI and analytical models for predictive maintenance, fault diagnostics, and performance monitoring.

Acts on processes

Automation

Any recurring on-site task as a workflow: alarm correlation, reporting, and exception routing.

Acts on equipment

Controls

Setpoint optimisation, dispatch, and safety actions executed locally, with approval gates.

03Product

All-in-one on-site data infrastructure.

Three modules, built on a canonical asset model, a multi-signal inference engine, and a containerized runtime that runs on standard industrial edge gateways.

Module 01 · Data unification

Automated data modeling & unification

  • Read-only discoveryConnects to installed SCADA, historians, OPC-UA servers, and gateways. Nothing is written to the control network.
  • Multi-signal tag inferenceProposes unit, asset class, role, and hierarchy for each tag from its name, behavior, prior validated sites, and neighbouring tags.
  • Confidence-tiered reviewHigh-confidence proposals are approved in bulk; uncertain ones go to a person with the reasoning shown.
  • Canonical asset modelOne versioned model of every asset and tag across all sites, with lineage to the raw tag name.
Map & contextualizeChisholm Ridge Solarcanonical-model v4.12
Tags1,984
Mapped1,572
In review0
Raw tagProposed meaningConf.
INV3_ACPWR_L1Inverter 3 / AC power / L1.94
PCS-02.kW_outPCS 02 / output power.91
TMP_CAB_03Cabinet 3 / temperature.86
BESS_07_SOCBattery 7 / state of charge.96
AI_0042Inverter 4 / phase L1.51
MET_GHIMet station / irradiance.92
Module 02 · On-site runtime

Optimized on-site execution

  • Purpose-built runtimeA headless engine with no UI, configuration tools, or local database, for large tag counts on standard gateways.
  • Model-bound acquisitionPer-tag rates across Modbus, OPC-UA, DNP3, S7 and others, each reading mapped to its canonical meaning.
  • Local feature engineeringRaw high-frequency signals become model inputs on the node; only summaries go upstream.
  • Isolated executionEach application runs in its own sandbox and continues through connectivity loss.
nr-gw-01 · North Ridgeruntime 1.8.4
Acquisition
Modbus
100 ms
OPC-UA
1 s
DNP3
250 ms
Sandboxed analytics
inverter.fault_predictor
bess.imbalance_monitor
alarm.correlator
Uplink down · buffering locallyanalytics running
Module 03 · Fleet orchestration

Fleet management & orchestration

  • Node registryLive inventory of every node: identity, connectivity, and versions of configuration, software, models, and certificates.
  • Model-addressed deploymentA change is expressed once against the canonical model and resolved to every affected node.
  • Staged rollout, automatic rollbackCanary waves with health checks at each stage. Offline nodes converge on reconnect.
  • Change control and auditEvery deployment carries an author, approvals, diff, and result.
Deployment campaignalarm.correlator v2.1.064 sites
Build
Canary
Wave 1
Fleet
deployedrolling outoffline · queued
Health checks · pending0 / 64 sites
author j.okafor · approvals 2/2diff +12 −3 · auto-rollback on

04Today vs. Packet

One data layer for every application.

LayerTodayWith Packet

Data unification

Mapped by hand

Each tag is mapped and contextualized manually, site by site.

Automated data modelling & setup

Sites come online fast.

Application execution

Rebuilt for every site

Re-architected for each hardware target. No reuse across sites.

Run models where the data is produced

New classes of use case viable.

Site operations

One node at a time

Patching, certificates, and updates handled individually, so cost grows with the fleet.

Single control plane for the fleet

Fleet growth without labour growth.

Deployment per site

3 months
1 week

05Architecture

Packet sits between your site systems and your applications.

Installed on site on standard industrial gateways. Deployed by Packet's Forward Deployed Engineers or in-house.

  1. 01 Connect

    Read-only ingestion

    From installed SCADA, historians, and OPC-UA servers.

  2. 02 Model

    AI-proposed tag meaning

    With human validation on every proposal.

  3. 03 Deploy

    Fleet-wide rollout

    Staged, with health checks and automatic rollback.

  4. 04 Run

    Local execution

    Applications run on site and continue through connectivity loss.